7 The J_ToBI Model of Japanese Intonation Jennifer J. Venditti 7.1. INTRODUCTION This chapter presents an overview of Japanese intonational structure and the transcription of this structure using J_ToBI,a variant of the general ToBI tagging scheme developed for Tokyo Japanese. Since the `Japanese ToBI Labelling Guidelines' (Venditti 1995) were ®rst distributed,J_ToBI has been used in numerous linguistic and computational contexts as a way to represent the intonation patterns of Japanese utterances. This chapter is intended not as a mere rehashing of the 1995 Guidelines,but rather as a comprehensive discussion of the fundamentals of Japanese intonation and the principles underlying the J_ToBI system. In Section 7.2,we describe the prosodic organization of Japanese and its intonational patterns.1 We discuss Japanese prosody from a cross-linguistic perspective,highlighting similarities between Japanese and other languages. Section 7.3 then provides an overview of the J_ToBI system. The discussion assumes the reader has some familiarity with intonation description,and with the general ToBI framework. Section 7.4 points out the differences between this new system and its predecessor,the Beckman±Pierrehumbert model presented in Japanese Tone Structure (Pierrehumbert and Beckman 1988). Section 7.5 gives an overview of the efforts toward automatization The author would like to thank Mary Beckman,Sun-Ah Jun,and Kikuo Maekawa for insightful discussions and comments throughout the ongoing development of the Japanese ToBI system. 1 This discussion and the J_ToBI system itself rely heavily on the model of Japanese tone structure put forth by Beckman and Pierrehumbert (see Beckman and Pierrehumbert 1986; Pierrehumbert and Beckman 1988, inter alia),which uses a tone-sequence approach to intonation modelling. However,a few important differences between J_ToBI and the Beckman±Pierrehumbert model will be discussed in Section 7.4. This approach is distinct from the superposition-based models of Japanese intonation (e.g. Fujisaki and Sudo 1971; Fujisaki and Hirose 1984; Venditti and van Santen 2000),which will not be discussed here. The J_ToBI Model of Japanese Intonation 173 of J_ToBI labelling,as well as the degree of labeller agreement using this system,and Section 7.6 lays out future directions for research on Japanese intonation. 7.2. JAPANESE PROSODIC ORGANIZATION AND INTONATION PATTERNS 7.2.1. Pitch accents Japanese is considered a pitch accent language,in that the intonational system uses pitch to mark certain syllables in the speech stream. In this way it is similar to languages like English,which also uses pitch accents in its into- national system. However,there are several fundamental differences between the two. First,Japanese and English differ in the level (lexical vs. post-lexical) at which pitch accent comes into play. In Japanese,pitch accent is a lexical property of a word,and thus the presence or absence of an accent on a particular syllable in a Japanese utterance can be predicted simply by knowing what word is being uttered. Take for example the minimal pair shown in Figure 7.1. Here,the verb /ueru/ in the phrase uerumono `something to plant' is lexically-speci®ed as unaccented,while that in ue'rumono `the ones who are UNACC ACC u e r u m o n o u e’ r u m o n o Figure 7.1 Waveforms and F0 contours of unaccented uerumono `something to plant' (left) and accented ue'rumono `the ones who are starved' (right) phrases, uttered by the same speaker. The x-axis represents the time-course of the utterances; the y-axis shows the frequency (in Hz) of the F0 contour. Both panels are plotted on the same frequency scale,and vertical lines mark the end of the second mora in each phrase. 174 Prosodic Typology starved' is speci®ed as accented on the second mora /e/.2 The accented phrase displays a precipitous fall in pitch starting near the end of this accented mora, while the unaccented phrase lacks such a fall.3 This lexical distinction con- trasts with languages such as English,in which pitch accents play a role at an entirely different level. In English,the location of metrically strong syllables in a word is determined at the lexical level,and it is these syllables (most often the strongest,or `primary-stressed' syllable) which serve as docking sites to which pitch accents may be associated at a post-lexical level. A second difference between the two languages is the function and dis- tribution of pitch accents. In English,pitch accents serve to highlight (or make `prominent') certain words or syllables in the discourse,and the distribution of pitch accents in an English utterance re¯ects this function. In a given utterance,there will be a number of metrically strong syllables that can potentially be made even more prominent by the association of a pitch accent. On which of these syllables pitch accents will fall is highly dependent on the linguistic structure of the utterance. That is,an interaction of various factors related to the syntax,semantics,pragmatics,discourse structure,attentional state,etc. will determine where the pitch accents are to be placed in English. In Japanese,in contrast,pitch accents are a lexical property of a given word,and thus they lack any such prominence-lending function. This leaves little room for variability in distribution of accents in a Japanese utterance. A third difference between the languages is the shapes and meanings of the pitch accents themselves. In Japanese there is only one type of pitch accent: a sharp fall from a high occurring near the end of the accented mora to a low in the following mora. In English,the inventory of pitch accent shapes is far more diverse. There are a number of pitch accent shapes,in which the F 0 can rise or fall to/from the accented syllable,or can maintain a local maximum/ minimum on that syllable. Each shape has associated with it a speci®c pragmatic meaning which that accent lends to the overall meaning of the intoned utterance (see e.g. Pierrehumbert and Hirschberg 1990). The Japanese falling accent does not have any such meaning associated with it. In summary,although both Japanese and English use pitch accent in their intonation system,the languages are in fact quite different with respect to the role that pitch accents play. The languages differ in the level at which pitch 2 In the transcriptions,accented words contain an apostrophe after the vowel with which the accentual fall is associated; unaccented words lack such a marking. 3 In the ®gure,the high to which the F 0 rises in the accented case (right) is higher than that in the unaccented case (left). This systematic height difference has been reported in previous studies (e.g. Poser 1984; Pierrehumbert and Beckman 1988; and many others). However,while accented peaks do tend to be higher than unaccented peaks,there is a large amount of variability in both,and there are plenty of cases in read and spontaneous speech where this relative height relation is reversed. Future The J_ToBI Model of Japanese Intonation 175 accents come into play,in the function and distribution of accents,and in the shapes and meanings of the accents in the inventory. 7.2.2. Prosodic groupings In addition to pitch accents,another important part of Japanese intonation is the grouping of words into prosodic phrases. Speakers can organize their speech into groups of intonational units,which are de®ned both tonally and by the degree of perceived disjuncture among words within/between groups. This grouping occurs at two levels in Japanese. First,there is a lower-level grouping,such as that shown in each panel in Figure 7.1. The verb ueru/ue'ru is combined with the following unaccented noun mono `thing or person',into a single prosodic phrase. This level of prosodic phrasing in Japanese is termed the accentual phrase (AP),and is typically characterized by a rise to a high around the second mora,and sub- sequent gradual fall to a low at the right edge of the phrase. This delimitative tonal pattern is a marking of the prosodic grouping itself,separate from the contribution of a pitch accent. Both panels in Figure 7.1 consist of a single accentual phrase with the delimitative tonal pattern,though the accented case (right panel) also shows the fall of the lexical accent.4 The degree of perceived disjuncture between words within an accentual phrase is less than that between sequential words with an accentual phrase boundary intervening. In Tokyo Japanese it is most common for unaccented words to combine with adjacent words to form accentual phrases,though under some circumstances a sequence of accented words may combine,in which case the leftmost accent survives and subsequent accents in the phrase are deleted. The second type of prosodic grouping in Japanese is the higher-level into- nation phrase (IP),which consists of a string of one or more accentual phrases. Like accentual phrases,this level of phrasing is also de®ned both tonally and by the degree of perceived disjuncture within/between the groups. However,the tonal markings and the degree of disjuncture for the IP are different from those of the accentual phrase. The intonation phrase is the prosodic domain within which pitch range is speci®ed,and thus at the start of each new phrase,the speaker chooses a new range which is independent of the former speci®cation. Since there also is a process of downstep in Japanese,by which the local pitch height of each accentual phrase is reduced when following a lexically accented investigations using large amounts of J_ToBI-tagged data are necessary in order to uncover the linguistic factors that are at work in determining this height relationship.
Details
-
File Typepdf
-
Upload Time-
-
Content LanguagesEnglish
-
Upload UserAnonymous/Not logged-in
-
File Pages29 Page
-
File Size-